st: Why does the full sample change when I add variables to the 2nd stageof a Heckman model?

[Gina Dokko <gdokko@stern.nyu.edu>]

I'm using a Heckman sample selection model for job performance 
(supinnov).  My selection model has 545 full observations, and includes 
both applicants and hired workers.  However, when I add certain 
variables to my main 2nd stage model, the total number of observations 
changes.  Below, I have a baseline model, followed by a model with 
variables from another source added (ctie csex cwehet density netsize).  
There are randomly missing observations for the other source, so the 
uncensored observations drops from 198 to 110.  However, the total 
number of obs. also drops by the same amount (545 to 457=88).  Why does 
the total number of obs. change, rather than the censored observations?  
Doesn't the total number of observations reflect the sample of the 
selection model?  Thanks for any help you can give! -Gina
*************************
. heckman supinnov recreferral tminjob tminjobsq ed_yrs2 file_age female 
propdum casdum au
> todum if adj==1, rob nolog select(occprior file_age ed_yrs2 female 
recreferral urate)

Heckman selection model                         Number of obs      
=       545
(regression model with sample selection)        Censored obs       
=       347
                                               Uncensored obs     
=       198

                                               Wald chi2(9)       =     
10.94
Log pseudo-likelihood = -570.3193               Prob > chi2        =    
0.2799

------------------------------------------------------------------------------
            |               Robust
            |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
supinnov     |
recreferral |   .0426596   .1285928     0.33   0.740    -.2093777    
.2946968
    tminjob |   .0025826    .006251     0.41   0.679    -.0096691    
.0148344
  tminjobsq |  -3.09e-06   .0000263    -0.12   0.907    -.0000547    
.0000485
    ed_yrs2 |    .091542   .0506159     1.81   0.071    -.0076634    
.1907475
   file_age |   -.011273   .0082991    -1.36   0.174     -.027539     
.004993
     female |  -.2280173   .1644796    -1.39   0.166    -.5503915    
.0943568
    propdum |  -.4738588   .2356279    -2.01   0.044     -.935681   
-.0120366
     casdum |  -.5085815   .2335728    -2.18   0.029    -.9663757   
-.0507873
    autodum |  -.4332596   .2199736    -1.97   0.049       -.8644   
-.0021192
      _cons |   4.296037   .4553609     9.43   0.000     3.403546    
5.188528
-------------+----------------------------------------------------------------
select       |
   occprior |  -.0001413   .0011995    -0.12   0.906    -.0024922    
.0022097
   file_age |  -.0075152   .0064384    -1.17   0.243    -.0201343    
.0051039
    ed_yrs2 |  -.1194929   .0278781    -4.29   0.000     -.174133   
-.0648528
     female |   .4391214   .1178484     3.73   0.000     .2081427    
.6701001
recreferral |   .0512239   .0990166     0.52   0.605     -.142845    
.2452927
      urate |   .5519167   .0622925     8.86   0.000     .4298257    
.6740077
      _cons |  -2.043613   .3599999    -5.68   0.000      -2.7492   
-1.338027
-------------+----------------------------------------------------------------
    /athrho |  -.2368843   .3414492    -0.69   0.488    -.9061124    
.4323439
   /lnsigma |   -.143962   .0694754    -2.07   0.038    -.2801313   
-.0077927
-------------+----------------------------------------------------------------
        rho |  -.2325507   .3229837                     -.7192611    
.4072783
      sigma |   .8659206   .0601602                      .7556845    
.9922376
     lambda |  -.2013704   .2904771                     -.7706951    
.3679543
------------------------------------------------------------------------------
Wald test of indep. eqns. (rho = 0): chi2(1) =     0.48   Prob > chi2 = 
0.4878
------------------------------------------------------------------------------
***************************
. heckman supinnov ctie csex cwehet density netsize recreferral tminjob 
tminjobsq ed_yrs2
> file_age female propdum casdum autodum if adj==1, rob nolog 
select(occprior file_age ed_
> yrs2 female recreferral urate)

Heckman selection model                         Number of obs      
=       457
(regression model with sample selection)        Censored obs       
=       347
                                               Uncensored obs     
=       110

                                               Wald chi2(14)      =     
60.00
Log pseudo-likelihood = -352.3372               Prob > chi2        =    
0.0000

------------------------------------------------------------------------------
            |               Robust
            |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
supinnov     |
       ctie |   1.031045   .2959833     3.48   0.000     .4509282    
1.611161
       csex |   .2903533   .1748974     1.66   0.097    -.0524394    
.6331459
     cwehet |   1.556572   .4502231     3.46   0.001     .6741512    
2.438993
    density |   .6961823   .7550956     0.92   0.357    -.7837779    
2.176142
    netsize |  -.0698929   .0371793    -1.88   0.060    -.1427629    
.0029772
recreferral |  -.1106137   .1566829    -0.71   0.480    -.4177065    
.1964791
    tminjob |  -.0181013   .0080461    -2.25   0.024    -.0338713   
-.0023312
  tminjobsq |   .0000884   .0000307     2.88   0.004     .0000283    
.0001485
    ed_yrs2 |   .2882946   .0565787     5.10   0.000     .1774023    
.3991869
   file_age |   .0011888   .0116109     0.10   0.918    -.0215682    
.0239459
     female |  -.0712856    .165677    -0.43   0.667    -.3960065    
.2534354
    propdum |  -.6029529   .2657829    -2.27   0.023    -1.123878    
-.082028
     casdum |  -.4712946   .2670187    -1.77   0.078    -.9946415    
.0520524
    autodum |  -.4085019   .2482766    -1.65   0.100    -.8951152    
.0781113
      _cons |   4.232665   .9127007     4.64   0.000     2.443804    
6.021526
-------------+----------------------------------------------------------------
select       |
   occprior |   -.004678     .00169    -2.77   0.006    -.0079903   
-.0013656
   file_age |   -.010334   .0071836    -1.44   0.150    -.0244136    
.0037456
    ed_yrs2 |  -.1120664   .0327407    -3.42   0.001    -.1762371   
-.0478958
     female |   .3216321    .136123     2.36   0.018     .0548359    
.5884284
recreferral |   .0707672   .1168943     0.61   0.545    -.1583414    
.2998757
      urate |   .4512408   .0710121     6.35   0.000     .3120596     
.590422
      _cons |  -1.782074   .3935421    -4.53   0.000    -2.553403   
-1.010746
-------------+----------------------------------------------------------------
    /athrho |  -.8558804   .2482602    -3.45   0.001    -1.342461   
-.3692994
   /lnsigma |  -.1011035   .1465839    -0.69   0.490    -.3884028    
.1861957
-------------+----------------------------------------------------------------
        rho |  -.6941291   .1286447                     -.8722622   
-.3533788
      sigma |   .9038395   .1324883                      .6781392    
1.204658
     lambda |  -.6273812   .2010785                     -1.021488   
-.2332746
------------------------------------------------------------------------------
Wald test of indep. eqns. (rho = 0): chi2(1) =    11.89   Prob > chi2 = 
0.0006
------------------------------------------------------------------------------ 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/